Accurate prediction of piperine content in black pepper using combined CNN and regression modelling with PDMAM@G electrode and cyclic voltammetry

Abstract

A novel graphite electrode with molecular imprints was developed for the selective and sensitive detection of piperine in black pepper. The electrode incorporates molecularly imprinted polymer (MIP) layers synthesized using poly (N,N-dimethylacrylamide) (PDMAM) as the monomer, ethylene glycol dimethacrylate (EGDMA) as the cross-linker, and piperine as the template, enabling specific recognition and quantification of piperine. Cyclic voltammetry (CV) was employed for electrochemical measurements, and the sensor was validated on black pepper samples from four different brands, demonstrating its practical applicability. To enhance prediction accuracy, convolutional neural network (CNN)-based feature extraction was combined with regression models for the analysis of CV signals. This hybrid approach, integrating CNN-extracted features with regression techniques such as K-nearest neighbour regressor (KNNR), gradient boost regressor (GBR), and random forest regressor (RFR), exhibited significant improvements in accuracy compared to the CNN model alone. Comprehensive experimental evaluations revealed that the CNN-KNNR model achieved a mean absolute percentage error of 0.034 and an R² value of 0.9999 when compared to reference values obtained through reverse-phase high-performance liquid chromatography (RP-HPLC).